Three major approaches have been taken to define non-classical multidrug resistance in cancer. In the first, we isolated KB cells (a subclone of HeLa) resistant to increasing levels of cisplatin (CP-r) and demonstrated multidrug resistance to arsenite and cadmium, to methotrexate, and to nucleoside analogs. This cross-resistance pattern is due to reduced uptake of each of these agents because their receptors have been relocalized from the cell surface into the cytoplasm of the cell. This relocalization of surface transporters appears to be due to altered recycling of these transporters due to alterations in the cytoskeleton that affect endocytic recycling compartments in cisplatin-resistant cells. Overexpression of the negative transcription regulator GCF2 occurs in cisplatin-resistant lines, which reduces expression of rhoA, causing disruption of the cytoskeleton as a proximate cause of this recycling defect. One additional consequence of reduced cell surface transporters is a reduction in glucose uptake and altered mitochondrial metabolism mediated by SIRT1. These changes are best understood as regulatory alterations due to epigenetic changes such as DNA methylation, histone modifications, and miRNA perturbations. The protein metallotheinein, heat shock proteins, ribosomal proteins, a selenoprotein, and the trans-membrane protein TMEM205 have been shown to play a role in cisplatin resistance. Expression of TMEM205, a membrane protein expressed in normal secretory cells, in combination with the small GTPase Rab8, confers cisplatin resistance. We have demonstrated changes in specific microRNAs (miRNAs), such as miRNA-181, consistently seen in cisplatin-resistant KB cells, and their contribution to drug resistance has been demonstrated by expression of miRNA mimics and inhibitors. In addition, a high throughput analysis of miRNAs that reverse the cisplatin resistance of KB-CP-r cells has identified WEE1 and CHK1 as essential elements of resistance to cisplatin. miRNA155 and miR-15 family members are miRNAs whose expression affects cisplatin resistance through WEE1 and CHK1. A second approach is to evaluate the unique features of melanoma cells that contribute to multidrug-resistance. One obvious feature of melanoma cells is the melanosome, a lysosome-derived organelle in which pigment formation takes place. We have shown that cisplatin is sequestered in this organelle, independent of extent of melanin formation, and extruded with melanosomes into the medium, reducing nuclear accumulation of this anti-cancer drug. Studies are underway to determine whether ABCB5, a transporter homologous to ABCB1, expressed at high levels in pigmented cells such as melanocytes and melanomas, contributes to the melanosomal sequestration seen in melanomas. Full-length ABCB5 has been expressed in KB cells, where it confers a broad multidrug resistance phenotype. ABCB5 knock-out mice have been generated, and they are viable, but have a metabolic alteration and a neurological disorder. In another approach, we have developed a Taqman Low Density Array (TLDA) microfluidic chip to detect mRNA expression of 380 different putative drug resistance genes and demonstrated that it is a sensitive, accurate, reproducible, and robust way to measure mRNA levels in tumor samples. Previous work from our laboratory indicates that mRNA measurements of levels of drug-resistance genes are, to a first approximation, predictive of functional expression of drug-resistance mechanisms. This drug-resistance chip has been applied to analysis of human cancers. One result from this analysis is that existing cancer cell lines do not mimic the expression patterns of actual human cancers for the 380 putative drug resistance genes chosen for the TLDA analysis and the simple expedient of growing cells in 3D culture does not correct this problem. This suggests the need for better in vitro cancer cell models to study multidrug resistance. Another conclusion is that a signature of eleven MDR genes we have studied predicts poor response in non-effusion ovarian cancer, and different subsets of 18 MDR genes predict poor response in ovarian cancer with effusions. For hepatoma, two different MDR gene expression signatures are associated with poor prognosis and better prognosis hepatoma. For acute myeloid leukemia (AML), recurrence of disease after remission induced by chemotherapy is associated with multiple different patterns of MDR gene expression, suggesting that for AML acquired resistance may be multifactorial. Validation of these results, indicating that MDR is complex and multifactorial in clinical cancers, will require the development of reliable in vitro culture models, and interpretation of these data using mathematical models based on network theory is proceeding. Towards this goal, we have developed a bioreactor that mimics capillary delivery (through silicon hydrogels) of oxygen to cells grown in 3D suspension. We have demonstrated physiological oxygen gradients and altered growth of cancer cells more closely approximating in vivo phenotypes. This bioreactor can be scaled up for growth of multiple cultures of primary cancer cells or cultured cancer cells to determine whether growth conditions play a primary role in affecting patterns of drug resistance. Because of the apparent complexity of drug resistance mechanisms in vivo, we have re-examined existing mathematical models that predict the development of drug resistance based on more homogeneous systems. In collaboration with Doron Levy (University of Maryland), we have formulated a new mathematical model that takes into account tumor heterogeneity and other features associated with in vivo systems.